We identified a strong pain point in UK's fare system, which causes 2 common problems for people wanting to buy fares online in advance:
- Purchasing an off-peak ticket: Customers buying fares lack the confidence that a cheaper fare option will allow them enough flexibility with their schedule, due to missing peak info/feedback during the purchasing process. This leads them missing out on better deals just to be safe.
- Traveling on an off-peak ticket: Customers who have already purchased a fare don’t have an easy way of finding the journeys they’ll most likely be able to choose on a given day in the future
We believe a large portion of users play it safe when selecting their fares due to a lack of insight in the chances of off-peak times for their future journey. By means of simple data-backed suggestions during both fare purchase and when determining their schedule, we aim to give users the confidence they need to always pick the optimal price.
How we built it
We ran time series analysis on the historical train journey data set provided by Trainline, and visualised the data in a dashboard to get insights in the off-peak trends. Using a trained time series forecasting model, Peak can predict for future dates the likeliness of catching an off-peak journey at any point in time.
We then built a progressive app for fare purchase, which enriches the provided options with suggestions for off-peak journeys.
Data processing was performed using Scala (Spark) and Python, and the results served to both the app and dashboard via a Node.js Express server.
Challenges we ran into
A technical challenge was finding the right parameters when running the analysis algorithms so that the results are meaningful and accurate. However, the real challenge was actually understanding the problem users faced, and finding out how to best present the information to the user to help them make better choices
Accomplishments that we're proud of
We built a sexy mockup + a prototype that actually works.
What we learned
The UK's railway system is a mess..
What's next for Peak
Enriching our model with more data, to provide even better suggestions to our users. But first: Sleep!